Integrating Binary Similarity Measures in the Link Prediction Task

Abstract

In this work we investigate the applicability of binary similarity and distance measures in the context of Link Prediction. Neighbourhood-based similarity measures to assess the similarity of nodes in a network have been long available. They boast the main advantage of low calculation complexity, because only a local view of the network is required. Neighbourhood-based measures are used in a variety of Link Prediction applications, including bioinformatics, bibliographic networks and recommender systems. It is possible to use binary measures in the same context, retaining the same prerogatives and possibly increasing the link prediction performances in domain-specific tasks. Preliminary studies have also been conducted on widely-accepted data sets.